Prefabricated beam steel strand stress detection method and system based on big data

Abstract

The application provides a prefabricated beam steel strand stress detection method and system based on big data. A two-dimensional sensor array is arranged to collect beam body vibration acceleration signals under natural multi-lane mixed driving conditions. A vehicle flow layering decoupling and blind source separation algorithm with simply supported beam boundary constraints are combined to remove multi-vehicle coupling and eccentric load interference and purify the beam body inherent vibration response. A standardized stress correlation model containing multiple correction coefficients is constructed. The beam type correction coefficient is iteratively calibrated based on the gradient descent algorithm with engineering constraints to accurately solve the steel strand stress of different beam types. A full bridge beam piece-sensor correlation database is built to carry out normal automatic detection of the whole region. The model is continuously iteratively optimized through the measured data backflow, and a multi-level abnormal early warning mechanism is used. The application does not need to close the road and control the vehicle, has strong anti-interference ability, good universality, high detection accuracy, can realize unmanned long-term monitoring of the whole region, and effectively guarantees the service safety of the prefabricated beam of the bridge.

Description

Technical Field

[0001] This invention belongs to the field of precast beam monitoring technology, specifically relating to a method and system for detecting stress in precast beam steel strands based on big data. Background Technology

[0002] Precast beams are the core load-bearing components of bridge engineering. Steel strands, as the main prestressed load-bearing structure of precast beams, directly determine the structural stability, load-bearing safety, and service life of the precast beams due to their actual stress state. During the long-term service of bridges, steel strands are susceptible to stress relaxation and stress exceeding limits due to factors such as environmental temperature and humidity, vehicle reciprocating loads, and structural aging. This can lead to cracking, deformation, and even structural failure of the precast beams. Therefore, accurate and real-time monitoring of steel strand stress is a core task in bridge health monitoring.

[0003] Currently, there are many shortcomings in the precast beam steel strand stress detection technology. The existing detection methods are mainly divided into two categories: manual static detection and real-time monitoring by fixed-point sensors. Neither of them can meet the long-term, high-precision monitoring requirements of bridges.

[0004] Manual static testing requires specialized testing equipment to inspect each beam and each point, which can only achieve periodic sampling inspection. The testing efficiency is extremely low, the labor cost is high, and it cannot capture the instantaneous stress changes of the steel strands under the dynamic load of vehicles. The test data is seriously lagging and cannot reflect the true service stress state of the precast beam.

[0005] Conventional fixed-point sensor monitoring often uses stress sensors directly attached to steel strands for detection. However, the sensors are cumbersome to install, easily damaged or shifted due to bridge construction, environmental corrosion, and structural deformation. They also have poor stability and durability, and can only collect local single-point stress data. They cannot be adapted to different precast beam models with different beam lengths and heights, resulting in extremely poor versatility.

[0006] In addition, existing technologies have not established a correlation mapping relationship between load vibration signals and steel strand stress, making it impossible to achieve indirect stress detection through non-contact, non-destructive vibration monitoring. Furthermore, the lack of a standardized beam type database and automated detection system results in large data dispersion and significant differences in detection accuracy for different beam types. This makes it impossible to achieve unified, routine, and accurate monitoring of all precast beams across the entire bridge, and makes it difficult to predict structural safety hazards caused by abnormal steel strand stress in advance. Summary of the Invention

[0007] To address the technical problems existing in the prior art, the first aspect of this invention is to provide a method for detecting the stress of precast beam steel strands based on big data. Secondly, based on the same inventive concept, this invention also provides a detection system for the aforementioned method for detecting the stress of precast beam steel strands based on big data. This system relies on big data modeling and dynamic vibration acceleration signals to construct a standardized stress correlation model and a full-bridge segment-sensor correlation database, thereby achieving non-destructive, real-time, and comprehensive detection of the stress of steel strands in multiple types of precast beams.

[0008] In this embodiment of the invention, the method for detecting stress in precast beam steel strands based on big data includes the following steps:

[0009] S1. Multiple sets of precast beams of different models with known reference stress of steel strands were selected as sample beams. Accelerometers were deployed on the surface of each sample beam in a two-dimensional sensor array covering the entire bridge lane to capture the vibration time delay differences of multi-lane vehicle excitation. Real-time acceleration signals of each sample beam were collected under the dynamic load condition of natural multi-lane mixed traffic flow without road closures or vehicle restrictions. The collected mixed traffic flow acceleration signals were processed in layers and decoupled. The lateral and longitudinal vibration time delay differences were used to distinguish the lane position and off-center load state of the vehicles. The multi-vehicle coupled vibration signals were decomposed by blind source separation with simple beam boundary constraints to obtain the pure beam's inherent vibration response. Combining the beam length and height structural parameters of each sample beam with the beam type correction coefficient, a standardized stress correlation model of precast beam steel strand stress, dynamic load acceleration signal, and beam structural parameters was constructed using a big data fitting algorithm.

[0010] S2. For all precast beams of the target bridge, accelerometers are installed on each precast beam to establish a bridge-sensor association database. The database includes the unique number of each precast beam, beam structure parameters, sensor placement location, sensor number, and calibrated beam type correction coefficient and lane eccentricity correction coefficient.

[0011] S3. Conduct routine vehicle dynamic load tests on the target bridge according to the preset time period. Directly use the daily multi-lane random mixed traffic flow of the bridge as excitation. There is no need to control vehicle speed, load and lane. Simultaneously collect the real-time acceleration signal of each precast beam of the entire bridge. Reuse the traffic flow layer decoupling process in step S1 to filter out lane load and multi-vehicle coupling interference. Import the collected and decoupled acceleration data and corresponding beam structure parameters into the standardized stress correlation model. Calculate the stress value of the steel strand of each precast beam in real time to complete the full-domain detection of the stress of the steel strand of the precast beam of the entire bridge.

[0012] The method of this invention, a precast beam steel strand stress detection system based on big data, includes a signal acquisition module, a data storage module, a model building module, a stress calculation module, and a timing control module. The signal acquisition module includes a two-dimensional acceleration sensor array deployed on each precast beam, used to acquire real-time vibration acceleration signals of the beam under multi-lane mixed traffic excitation, and to perform raw data preprocessing and uploading. The data storage module stores beam structural parameters, beam type correction coefficients, lane eccentricity correction coefficients, a full bridge segment-sensor association database, and historical detection data for each type of precast beam. The model building module incorporates a signal decoupling unit and iterative optimization. The unit is used to perform layered decoupling and blind source separation of the original vibration signal to purify the inherent response of the beam. Based on sample data, iterative calibration correction coefficients are determined to construct a standardized stress correlation model. The stress calculation module is bidirectionally connected to the signal acquisition module, model construction module, and data storage module. It is used to receive the purified acceleration data, retrieve model parameters and corresponding beam structural parameters, calculate the stress value of the steel strand in real time through the standardized stress correlation model, and feed the measured data back to the model construction module to achieve iterative optimization of the model. The timing control module is used to automatically trigger signal acquisition and detection operations according to a preset cycle to realize normalized automatic detection of steel strand stress.

[0013] Compared with the prior art, the advantages of the preferred technical solution of the present invention include:

[0014] 1. This invention relies on a two-dimensional sensor array for data acquisition, vehicle flow layer decoupling, and a blind source separation algorithm with simply supported beam boundary constraints. It effectively eliminates detection interference caused by multi-vehicle coupled vibration and lane eccentric loading, overcoming the limitations of traditional detection methods that require road closures and vehicle control, and reliance on standard test conditions. At the same time, it combines a standardized stress correlation model with a full-bridge segment-sensor correlation database to achieve non-destructive, dynamic, and routine detection of the stress of steel strands in precast beams of different specifications. It accurately captures real-time dynamic stress changes in steel strands, improves the detection's anti-interference capability, versatility, and adaptability, and solves the technical problems of low efficiency, poor accuracy, and inability to adapt to real bridge service conditions in traditional detection methods.

[0015] 2. This invention constructs a standardized stress correlation model, which uses beam type correction coefficient to offset differences in beam structure and lane eccentricity correction coefficient to compensate for traffic load interference, thus solving the technical defects of traditional models such as poor universality and limited adaptability to a single scenario, and achieving unified and accurate detection of multiple types of precast beams.

[0016] 3. This invention establishes a full-bridge segment-sensor association database, combined with a time-series automated control mechanism, to achieve unattended, routine, and blind-spot-free detection of the entire precast bridge beam. This solves the problems of low efficiency, data lag, and inability to capture dynamic stress changes in traditional manual sampling inspections, thereby reducing operation and maintenance costs.

[0017] 4. This invention establishes a detection system integrating signal acquisition, data storage, model construction, stress calculation, timing control, and early warning display. Each module works in synergy and supports reverse feedback of measured data to iteratively optimize model parameters, forming a fully automated closed-loop monitoring system of "acquisition-calculation-analysis-optimization-early warning". It can achieve unattended routine monitoring of precast bridge beam stress without manual intervention, and has the advantages of high automation, strong adaptive optimization capability, and no blind spots in monitoring coverage. It effectively ensures the stability and accuracy of long-term monitoring of bridge steel strand stress and provides reliable data support for the healthy operation and maintenance of bridge structures. Attached Figure Description

[0018] Figure 1 This is a flowchart of the big data-based stress detection method for precast beam steel strands, as described in this embodiment.

[0019] Figure 2 This is a schematic diagram of the structure of a precast beam steel strand stress detection system based on big data, as described in this embodiment. Detailed Implementation

[0020] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0021] Example 1

[0022] This embodiment provides a method for detecting the stress of precast beam steel strands based on big data, such as... Figure 1 As shown, in a preferred embodiment, the detection method includes the following steps:

[0023] S1. Modeling and Parameter Calibration: Multiple sets of precast beams of different types with known steel strand reference stresses were selected as sample beams. Accelerometers were deployed on the surface of each sample beam in a two-dimensional sensor array covering the entire bridge lanes to capture the vibration time delay differences caused by multi-lane vehicle excitation. The two-dimensional sensor array was evenly distributed laterally along each lane of the bridge and longitudinally along the span of the precast beams, with no less than 3 rows and 4 columns of sensor points per beam to ensure full coverage of multi-directional vibration signal acquisition.

[0024] Real-time acceleration signals of each sample beam were collected under the dynamic load condition of natural multi-lane mixed traffic flow without road closures or vehicle restrictions. The collected mixed traffic flow acceleration signals underwent hierarchical decoupling processing. The lateral and longitudinal vibration time delay differences were used to distinguish the vehicle's lane position and off-center load state. Blind source separation with simply supported beam boundary constraints decomposed the multi-vehicle coupled vibration signals, eliminating multi-vehicle coupling and lane off-center load interference to obtain the pure beam's inherent vibration response. Specifically, the traffic flow hierarchical decoupling involved setting a 0.01s time delay threshold and determining the vehicle's lane and lateral off-center load distance based on the vibration signal time difference at different sensor locations, dividing the coupled vibration signals into independent lane vibration components. The blind source separation algorithm introduced deflection constraint boundary conditions at both ends of the simply supported beam, iteratively eliminating vehicle load vibration components and environmental noise components, retaining only the beam structure's inherent vibration response. This embodiment uses the FastICA independent component analysis algorithm for blind source separation, using the maximum negative entropy as the objective loss function to achieve iterative convergence.

[0025] By combining the beam length and height structural parameters of each sample beam with beam type correction coefficients, a standardized stress correlation model of precast beam steel strand stress, dynamic load acceleration signal, and beam structural parameters is constructed using a big data fitting algorithm. Specifically, the big data fitting employs a multivariate nonlinear fitting method, using measured reference stress as the label and structural parameters and decoupled acceleration as features to obtain a standardized correlation model with multiple correction coefficients.

[0026] S2. Construction of a full bridge segment information database: For all precast beams of the target bridge, accelerometers are installed on each precast beam. The sensors are fixed by waterproof adhesive. The installation locations are uniformly recorded as standard points at mid-span, 1 / 4 span, and 3 / 4 span. All parameters and calibration coefficients are associated with each other and entered into the database and stored in encrypted form. A full bridge segment-sensor association database is established to achieve precise binding of beam, equipment, and model parameters. The database includes the unique number of each precast beam, beam structural parameters, sensor installation location, sensor number, and the calibrated beam type correction coefficient and lane eccentricity correction coefficient.

[0027] S3. Routine Dynamic Stress Testing: Routine vehicle dynamic load tests are conducted on the target bridge according to a preset time cycle (e.g., 12 hours / test for key highway bridges and 24 hours / test for ordinary municipal bridges). The daily multi-lane random mixed traffic flow of the bridge is directly used as the excitation, without the need to control vehicle speed, load and lanes. Real-time acceleration signals of each precast beam of the entire bridge are collected synchronously. The traffic flow layering decoupling process in step S1 is reused to filter out lane load and multi-vehicle coupling interference. The collected and decoupled acceleration data and corresponding beam structural parameters are imported into a standardized stress correlation model to calculate the stress value of the steel strands of each precast beam in real time, thus completing the full-domain detection of the stress of the steel strands of the precast beams of the entire bridge.

[0028] S4. Stress Anomaly Early Warning: The system pre-stores the safety stress threshold for each beam type. It compares the real-time calculated stress value of the steel strand with the preset safety stress threshold. When the calculated stress value of the steel strand exceeds the preset safety stress threshold range, an anomaly early warning message is generated. The message is then bound to the corresponding precast beam number, abnormal data, real-time stress value, threshold range, sensor location information, and detection time to generate a standardized anomaly early warning message. The message is uploaded to the operation and maintenance terminal in real time via wireless transmission and is simultaneously stored in the background log, enabling the anomaly data to be traceable and queryable.

[0029] The safety stress threshold is preset and dynamically updated based on the design parameters, service life, and environmental conditions of different beam types, enabling precise classification and judgment. Precast beams use the standard safety stress threshold indicated on the design drawings; for bridges with 5-10 years of service, the safety stress threshold is lowered by 5%; for bridges with over 10 years of service, located in coastal areas with high corrosion and extreme temperature differences, the safety stress threshold is lowered by 10%. The system automatically updates the threshold parameters quarterly based on bridge inspection reports to adapt to performance degradation caused by beam aging and environmental erosion, avoiding false or missed warnings.

[0030] In step S1 of this invention, the calculation formula for the standardized stress correlation model is as follows:

[0031]

[0032] in, The calculated stress value of the precast beam steel strands, in MPa, is the final output of the model, used to characterize the actual prestress state of the steel strands. This is a beam type correction coefficient, dimensionless, obtained through gradient descent iterative calibration, used to offset systematic deviations caused by different beam stiffness and cross-sectional properties; The lane eccentricity correction coefficient is dimensionless and is obtained through multi-condition off-center load test calibration. It is used to compensate for detection interference caused by vehicle driving in different lanes and lateral off-center load. A uniform calibration constant for mechanical dimensions, in MPa, is calibrated by a standard laboratory tension test and remains constant throughout the test. It is used to uniformly map dimensionless comprehensive characteristic quantities to stress dimensions. The length of the precast beam is in meters (m). The height of the precast beam is in meters (m). To express the natural vibration acceleration of the beam after decoupling, in units The vibration signal of the original traffic flow is obtained by collecting the original traffic flow vibration signal from the accelerometer and then decoupling it through blind source separation; Reference acceleration, in units The system uniformly calibrates reference values ​​for global normalization of acceleration data, ensuring the comparability of data collected in different time periods and environments; This is the unloaded reference acceleration, in units. The standard unloaded steady-state reference vibration value of the precast bridge beam is obtained by selecting the same type of precast beam under the conditions of no defects, no load, no wind and stable temperature, and collecting the steady-state vibration acceleration multiple times and taking the average value.

[0033] The standardized stress correlation model of this invention is a dedicated fitting model adapted to multi-beam and multi-traffic conditions. During calculation, data normalization preprocessing is uniformly performed, and the measured acceleration is used as the basis for calculation. respectively with , Perform ratio calculations to eliminate numerical deviations caused by sensor drift and ambient temperature and humidity; beam type correction coefficient. Lane eccentricity correction coefficient Each beam type is individually calibrated, categorized, and stored. During calculation, the corresponding beam type structural parameters are retrieved and substituted into the formula to complete the superposition calculation of linear and nonlinear correction terms, and output the actual stress value of the steel strand.

[0034] In step S1 of the present invention, the beam type correction coefficient The calibration method is as follows: For precast beams of different models, multiple sets of acceleration signal data corresponding to known reference stresses are collected. The least squares method combined with the gradient descent iterative optimization algorithm is used to complete the parameter solution and convergence calibration. During the iteration process, the upper and lower bound constraints of the steel strand specification stress are forcibly added, and only valid solutions that conform to the engineering mechanics range are retained. For example, for precast beams with different spans and cross-sectional types such as 20m, 25m, and 30m, no less than 20 sets of sample data are selected. Each set of samples includes the measured reference stress and the decoupled valid acceleration data. The least squares method is used to construct the fitting loss function, and the gradient descent algorithm is used to iteratively update the function. Numerical values; during the iteration process, stress upper and lower limits were strictly set according to the "Code for Design of Prestressed Concrete Bridges for Highways", invalid fitted solutions exceeding the code range were automatically eliminated, and the final value was locked after iteration convergence. The value is adjusted to ensure that the correction factor conforms to the actual mechanical laws of the engineering project.

[0035] Beam type correction factor The calibration method is as follows:

[0036] 1) Construct the target loss function for sample fitting:

[0037]

[0038] in, The value of the iterative loss function, in units. , used to characterize the fitting deviation between the calculated stress and the actual stress in the model; The number of precast beam groups used for the effective sample calibration of single beam type is dimensionless and manually preset. Usually, N≥20 is taken to ensure the robustness of the fit. The reference true stress of the steel strand of the precast beam in the i-th sample group is measured in MPa through tension calibration test or by third-party static testing equipment, and is used as the true value label. Let be the structural-acceleration comprehensive characteristic quantity of the precast beam in the i-th sample group, which is dimensionless and calculated using the following formula: All parameters are normalized and satisfy the rules of addition and subtraction. Let be the steady-state acceleration of the i-th group, in units ; , where is the beam type correction coefficient to be solved, is dimensionless, and is the beam type normalization correction parameter, which is only used to offset the system fitting deviation caused by different beam structures and stiffnesses;

[0039] Iterative constraints: ,in and The values ​​represent the minimum and maximum allowable stresses for the corresponding beam-type steel strands, in MPa, taken from bridge design specifications.

[0040] 2) Iteratively update the correction coefficients using gradient descent:

[0041]

[0042] in, The correction coefficient for the k-th iteration is dimensionless and represents a variable in the iteration process. is the correction coefficient after the (k+1)th iteration update, which is dimensionless; The iterative learning rate is derived in the formula using a closed-loop method. The hyperparameter is manually preset and fixed at a value of 0.001 to 0.01, which is used to control the iterative convergence speed and stability. The partial derivative of the loss function with respect to the correction coefficients represents the direction of the parameter gradient update. The partial derivative formula is as follows: The closed-loop derivation of the dimensions in the formula;

[0043] 3) Iterative convergence criteria:

[0044]

[0045] The left side of the formula represents the loss function relative to the iteration error, which is dimensionless and calculated from the loss values ​​of two consecutive iterations. When the relative error is consistently less than 0.5%, the model is considered converged, and the current iteration is locked. This is the final correction factor for the corresponding beam type.

[0046] For example, preset initial values ​​for the initial iterations during the initial calibration. =1, iterative learning rate The value is fixed at 0.005; the comprehensive feature quantity is calculated by substituting the sample data into each group. With loss function value The gradient direction is solved by the partial derivative formula and then iteratively updated. Numerical values; after each iteration, it automatically determines whether the engineering stress constraint conditions are met and removes abnormal data samples; when the relative error is less than 0.5% for 5 consecutive iterations, it is considered convergent, the iteration stops and the beam type is saved. Values; all beam type parameters are independently calibrated, classified and archived to avoid confusion between different beam type parameters and ensure the model fitting accuracy.

[0047] Example 2

[0048] This embodiment provides a big data-based precast beam steel strand stress detection system, which is used in the big data-based precast beam steel strand stress detection method of Embodiment 1. The system includes a signal acquisition module, a data storage module, a model building module, a stress calculation module, a timing control module, and a terminal early warning module.

[0049] The signal acquisition module includes a two-dimensional accelerometer array deployed on each precast beam. This array collects real-time vibration acceleration signals of the beams under multi-lane mixed traffic excitation and performs raw data preprocessing and uploading. Specifically, 1000Hz high-sampling-frequency piezoelectric accelerometers are used, featuring waterproof, dustproof, and anti-interference capabilities. The array is deployed to fully cover the beam surface. After acquiring the raw vibration signals, it automatically performs preliminary noise reduction and filtering preprocessing, and then uploads the data via 4G / 5G wireless communication.

[0050] The data storage module stores structural parameters, beam type correction coefficients, lane eccentricity correction coefficients, a full bridge segment-sensor association database, stress safety thresholds, and historical inspection data for various types of precast beams. The module employs a dual-storage model (cloud + local) to categorize and archive beam parameters, model coefficients, threshold data, and historical inspection records, supporting rapid data retrieval and traceability.

[0051] The model building module incorporates a signal decoupling unit and an iterative optimization unit. These units perform layered decoupling of the original vibration signals, blind source separation to purify the inherent response of the beam, and iteratively calibrate correction coefficients based on sample data to construct a standardized stress correlation model. The system automatically selects valid, anomaly-free measured stress data daily as iterative samples for input into the model building module. Monthly, it automatically initiates a gradient descent iterative optimization process, updating the beam type correction coefficients and lane eccentricity correction coefficients under specified stress constraints, and simultaneously updating the standardized stress correlation model. After iteration, the system automatically compares the accuracy of the old and new models, retaining the optimal model parameters. This enables continuous optimization of detection accuracy and adaptive adaptation to changes in bridge service conditions during long-term system operation.

[0052] The stress calculation module is bidirectionally connected to the signal acquisition module, model building module, and data storage module, serving as the core computing unit of the system. It interacts with each module in real time to achieve data exchange, stress calculation, and data feedback, forming a closed-loop optimization. The stress calculation module receives purified acceleration data, retrieves model parameters and corresponding beam structural parameters, calculates the stress value of the steel strands in real time using a standardized stress correlation model, and feeds the measured data back to the model building module for iterative model optimization, improving long-term detection accuracy.

[0053] The timing control module automatically triggers signal acquisition and detection operations according to a preset cycle, enabling routine automatic detection of steel strand stress. The module has a built-in timer program that automatically starts and stops the acquisition and calculation operations according to the preset cycle, requiring no manual operation.

[0054] The terminal early warning module receives the detection results from the stress calculation module, compares the calculated stress value of the steel strand with the preset safe stress threshold, determines the stress anomaly state, and completes stress anomaly determination, graded early warning, data display, and storage. The terminal early warning module supports graded early warning, setting three warning threshold ranges. Based on the degree of stress exceeding the limit, it is divided into general early warning, intermediate early warning, and advanced early warning, corresponding to the push of differentiated maintenance prompts and operation and maintenance suggestions. For example, stress exceeding the threshold by 0-10% triggers a general early warning, pushing a reminder for regular review; stress exceeding the threshold by 10%-20% triggers an intermediate early warning, pushing a reminder for timely maintenance; stress exceeding the threshold by more than 20% triggers an advanced early warning, with real-time pop-up alarms and SMS pushes to operation and maintenance personnel, requiring emergency investigation and handling, achieving risk-level control and precise operation and maintenance.

[0055] In this invention, the detection system also includes a data visualization module for real-time display of stress data, historical variation curves, and anomaly records for each precast beam, supporting data querying, exporting, and statistical analysis. The data visualization module builds a panoramic bridge monitoring interface, updating single-beam stress values ​​and daily / weekly / monthly variation curves in real-time using beam number as the dimension, automatically summarizing anomaly records and generating monitoring reports; it supports filtering and querying data by time, beam number, and warning level, and allows one-click export of reports and monitoring logs, providing data support for bridge health assessment and maintenance archiving.

[0056] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for detecting stress in precast beam steel strands based on big data, characterized in that, Includes the following steps: S1. Select multiple sets of precast beams of different models with known reference stress of steel strands as sample beams. Arrange acceleration sensors on the surface of each sample beam in the form of a two-dimensional sensor array covering the entire lane of the bridge to capture the vibration time delay difference of multi-lane vehicle excitation. Under the dynamic load condition of natural multi-lane mixed traffic flow without road closures or vehicle restrictions, collect the real-time acceleration signals of each sample beam. Use the lateral and longitudinal vibration time delay difference to distinguish the lane position and off-center load state of the vehicle. Decompose the multi-vehicle coupled vibration signal through blind source separation with simple beam boundary constraints to obtain the pure beam's inherent vibration response. Combine the beam structure parameters of beam length and beam height of each sample beam and the beam type correction coefficient, construct a standardized stress correlation model of precast beam steel strand stress, dynamic load acceleration signal, and beam structure parameters through big data fitting algorithm. S2. For all precast beams of the target bridge, an acceleration sensor is installed on each precast beam to establish a whole bridge beam-sensor association database. The database includes the unique number of each precast beam, beam structure parameters, sensor placement location, sensor number, and calibrated beam type correction coefficient and lane eccentricity correction coefficient. S3. Conduct routine vehicle dynamic load tests on the target bridge according to the preset time period. Directly use the daily multi-lane random mixed traffic flow of the bridge as excitation. There is no need to control vehicle speed, load and lane. Simultaneously collect the real-time acceleration signal of each precast beam of the entire bridge. Reuse the traffic flow layer decoupling process in step S1 to filter out lane load and multi-vehicle coupling interference. Import the collected and decoupled acceleration data and corresponding beam structure parameters into the standardized stress correlation model. Calculate the stress value of the steel strand of each precast beam in real time to complete the full-domain detection of the stress of the steel strand of the precast beam of the entire bridge.

2. The method for detecting stress in precast beam steel strands based on big data according to claim 1, characterized in that, In step S1, the calculation formula for the standardized stress correlation model is as follows: in, The model's final output, based on the calculated stress values ​​of the precast beam steel strands, characterizes the actual prestressing state of the steel strands. This is a beam type correction factor used to offset systematic deviations caused by different beam stiffness and section properties; This is the lane eccentricity correction coefficient, used to compensate for detection interference caused by vehicles traveling in different lanes and lateral load. A constant for the unified calibration of mechanical dimensions, used to uniformly map dimensionless comprehensive characteristic quantities into stress dimensions; The length of the precast beam; The height of the precast beam; This refers to the natural vibration acceleration of the beam after decoupling; It serves as a benchmark acceleration, used for global normalization of acceleration data to ensure the comparability of data collected at different times and in different environments; is the unloaded reference acceleration, and is the standard unloaded steady-state reference vibration value of the precast bridge beam.

3. The method for detecting stress in precast beam steel strands based on big data according to claim 2, characterized in that, In step S1, the beam type correction factor The calibration method is as follows: For different types of precast beams, multiple sets of acceleration signal data corresponding to known reference stresses are collected, and the least squares method combined with gradient descent iterative optimization algorithm is used to complete the parameter solution and convergence calibration; during the iteration process, the upper and lower bound constraints of the steel strand specification stress are forcibly added, and only the effective solutions that conform to the engineering mechanics range are retained.

4. The method for detecting stress in precast beam steel strands based on big data according to claim 3, characterized in that, The beam type correction factor The calibration method is as follows: 1) Construct the target loss function for sample fitting: in, The value of the iterative loss function is used to characterize the fitting deviation between the model's calculated stress and the actual stress. The number of valid sample precast beam groups used for single-beam calibration; The reference true stress of the steel strand in the precast beam of the i-th sample group; The structural-acceleration comprehensive characteristic of the precast beam in the i-th sample group is calculated using the following formula: All parameters are normalized and satisfy the rules of addition and subtraction. Let i be the steady-state acceleration sampled in the i-th group; is the beam type correction coefficient to be solved, and is the beam type normalization correction parameter, which is only used to offset the system fitting deviation caused by different beam structures and stiffnesses; Iterative constraints: ,in and The minimum and maximum allowable stresses are specified in the corresponding beam steel strand specifications; 2) Iteratively update the correction coefficients using gradient descent: in, The correction coefficient for the k-th iteration is a variable in the iteration process. These are the correction coefficients after the (k+1)th iteration update; The hyperparameters are manually preset as the learning rate to control the iteration convergence speed and stability. Let be the partial derivative of the loss function with respect to the correction coefficients, representing the direction of the parameter gradient update. The partial derivative formula is: ; 3) Iterative convergence criteria: The left side of the formula represents the relative iteration error of the loss function, calculated from the loss values ​​of two consecutive iterations. When the relative error is consistently less than 0.5%, the model is considered converged, and the current iteration is locked. This is the final correction factor for the corresponding beam type.

5. The method for detecting stress in precast beam steel strands based on big data according to any one of claims 1-4, characterized in that, After step S3 is completed, an abnormal stress warning step is also included: the real-time calculated stress value of the steel strand is compared with the preset safe stress threshold. When the calculated stress value of the steel strand exceeds the preset safe stress threshold range, an abnormal warning message is generated and uploaded to the terminal along with the corresponding precast beam number, abnormal data, and detection time.

6. The method for detecting stress in precast beam steel strands based on big data according to claim 5, characterized in that, The safety stress threshold is preset and dynamically updated based on the design parameters, service life, and environmental conditions of different beam types, so as to achieve accurate classification and judgment.

7. A precast beam steel strand stress detection system based on big data, used to execute the precast beam steel strand stress detection method according to any one of claims 1-6, characterized in that, It includes a signal acquisition module, a data storage module, a model building module, a stress calculation module, and a timing control module; The signal acquisition module includes a two-dimensional acceleration sensor array deployed on each precast beam, used to acquire real-time vibration acceleration signals of the beam under the excitation of multi-lane mixed traffic flow, and to complete the preprocessing and uploading of raw data. The data storage module is used to store the beam structure parameters, beam type correction coefficient, lane eccentricity correction coefficient, full bridge segment-sensor association database, and historical detection data of each type of precast beam; The model building module has a built-in signal decoupling unit and an iterative optimization unit, which are used to perform hierarchical decoupling of the original vibration signal, separate blind sources to purify the inherent response of the beam, and iteratively calibrate the correction coefficients based on sample data to build the standardized stress correlation model. The stress calculation module is bidirectionally connected to the signal acquisition module, the model building module, and the data storage module. It is used to receive the purified acceleration data, retrieve the model parameters and corresponding beam structure parameters, calculate the stress value of the steel strand in real time through the standardized stress correlation model, and feed the measured data back to the model building module to realize the iterative optimization of the model. The timing control module is used to automatically trigger signal acquisition and detection operations according to a preset cycle, so as to realize the automatic detection of stress in steel strands under normal conditions.

8. The precast beam steel strand stress detection system based on big data according to claim 7, characterized in that, It also includes a terminal early warning module, which receives the detection results from the stress calculation module, compares the calculated steel strand stress value with the preset safety stress threshold, determines the stress anomaly state, and completes stress anomaly determination, graded early warning, data display and storage.

9. The precast beam steel strand stress detection system based on big data according to claim 8, characterized in that, The terminal early warning module supports tiered early warning, which is divided into general early warning, intermediate early warning and advanced early warning according to the stress exceeding the limit, and pushes maintenance prompt information of different levels accordingly.

10. The precast beam steel strand stress detection system based on big data according to any one of claims 7-9, characterized in that, It also includes a data visualization module, which is used to display stress data, historical change curves, and anomaly records of each precast beam in real time.

Images